Identifying the Best Model to Explain Climate Change-Induced Migration

Kate Ricke, Gaurav Khanna, you are both 2023–24 recipients of IGCC’s Climate Change and Security grant. What’s the problem you’d like to solve?

Gaurav Khanna: Climate change is real and it’s going to have a huge impact on all of our lives. People think of adapting to climate in many different ways, such as changing aspects of their lifestyles, but the other big thing is that people are going to move.

Climate-induced migration is a complex problem. We haven’t yet seen the kind of climate change that we’re going to have in the future, so it’s difficult to grasp what the impacts might be. The climate is going to change in many different ways—not only will it mean hotter temperatures, but there are also going to be more droughts, floods, storm surges, and sea level rises, as well as more variability in climate and changes in precipitation patterns.

All of these affect people’s livelihoods differently. They might affect agricultural livelihoods, or they might cause direct displacement when people living near the coast are forced to move. There are also going to be different ways in which people adapt. Some can make stronger houses, but others who have the ability to move will do so. Their responses are also going to be different based on cost—some people might not be able to afford to uproot their entire livelihoods, so they might get trapped where they are. Or there might be forced displacement, when a flood forces people to move because their livelihoods get destroyed.

Climate migration is a very complex phenomenon because it interacts with income levels, different kinds of shocks—whether it’s about variability or just temperature changes—and regional factors. Kate has developed what we call a “modular framework” to try to really get at these complexities by dividing them up in different ways.

Kate Ricke: I came to this project as someone who has worked on trying to put together the pieces of physical projections of what’s happening with climate change, especially regionally, and how to couple that with simple models of social systems. Gaurav came with the knowledge of how migration actually works and plays out. This project with IGCC is an extension and new flavor of some work we’ve been doing for a long time, and we’ve just been trying to plug away at a problem that we knew was going to be hard and has turned out to be way harder than we had anticipated.

Is sea level rise-induced migration a newer intersection of research in this space? Or has there been significant work on this in the past?

Kate Ricke: It’s a topic of huge interest to people and is something that everyone has wanted to understand better, but it’s something I knew absolutely nothing about when we got started. The data that’s available about human migration is sparse compared to data we have about the climate. It’s in different timescales and resolutions, and the factors that contribute to human migration are so complex compared to some other equally important impacts of climate change. Migration is really interesting in that it’s hard to classify as either a climate change impact or a form of adaptation. It’s actually both, and depending on whose perspective you’re looking at, it would be classified differently. So, I don’t know that it’s new, but no one’s really cracked this problem open yet in terms of the right theoretical frameworks or approach to data.

Gaurav Khanna: We think our project is comprehensive in that it ticks a lot of important boxes. There’s been a lot of great work that focuses on a certain region, such as looking at displacement in the Bangladesh delta or heat-induced migration in rural Indonesia. We wanted to do a global analysis focused on a wider set of variables.

A lot of people have focused only on changes in temperature and precipitation. But temperature and precipitation might not be the only things that individuals base the decision to leave on. But more variability in temperature, or changes in the trend in temperature—where they see it’s getting hotter over the years and crops are going to fail—these are the things that might be what drive people to change. So, what’s new about our work is that we are trying to think about what these metrics of the change in climate are that might affect why people move.

The third thing is that a lot of people have documented these relationships between temperature and migration or precipitation and migration. We don’t want to just stop with these historical relationships, and instead want to make a statement about what’s going to happen in the future. To do that, we want to think carefully not just about where people live and how temperature and precipitation will change in different parts of the world, but also about how institutions might change in response to what’s going to happen in the future.

I want to go back to something you said, Kate, about how it’s hard to classify migration as a climate change impact or a form of adaptation. Could you please elaborate on that?

Kate Ricke: The standard framework for thinking about climate change’s effects on human welfare is to think about the problems first and then look for solutions. We first try to understand the physical system and the stressors that humans are applying to the Earth’s system in terms of how they change different Earth and environmental indicators such as temperature, precipitation, sea level rise, etc. We then take those environmental indicators and translate them into what people call impacts, so we do impact assessments. We ask what it means in terms of changes to crop yields, instances of heat stroke, increases in storm surges during tropical storms, etc. Those are what we call climate change impacts.

From that, we try and figure out how humans can respond to impacts in order to reduce suffering from climate change. These can be things that happen with or without policy intervention but are responses to those impacts to lessen how they’re negatively experienced. Those things are called adaptation.

Views on migration change depending on who you’re talking to, whether it’s decision-makers, climate scientists, or the general public. Some see it as an impact—migrants coming to the U.S. is an impact of climate change on the U.S.—whereas clearly migration is also a response to climate change impacts and a form of adaptation.

Are there features or consequences that differentiate sea level rise-induced migration from other triggers of migration, such as conflict or ethnic persecution?

Kate Ricke: Yes, I think so. The big challenge with sea level rise-induced migration is that it’s not as easy to find analogues from the recent past to apply. Like Gaurav said, we spent a lot of time looking at how migration might change as a response to changes in temperature and precipitation, and the approach we’ve taken is to look at data to see how temperature and precipitation changes due to natural variability in the climate system have induced migration responses. We can’t take the same approach with sea level rise because there hasn’t been sea level rise in the recent past that we can use as an analogue to what might happen in the future.

The [United Nations Framework Convention on Climate Change], the major climate international framework and treaty system for talking about climate change, classifies different types of climate-induced migration. What we’ve mostly been thinking about is adaptive migration, where people voluntarily relocate because conditions have changed in a way that makes it more advantageous to move than to stay. For sea level rise, it’s more likely to be other types of migration: involuntary displacement, which is a response to disasters, or organized relocation, which is something that people are talking about a lot with sea level rise. Depending on the type of migration, where people might go will vary quite a lot; organized relocation is going to look very different than involuntary displacement or adaptive migration. That’s a new challenge we’re facing when we start thinking about sea level rise.

Gaurav Khanna: With sea level rise, there’s this very stark visual of displacement. It’s not necessarily a slow-moving thing where people notice, for example, that there have been regular droughts over the years and so the crops are doing badly. It’s more apparent, since it means people’s houses are going to be underwater. Especially in certain island nations, it’s going to change entire livelihoods. What little discussion there has been on the institutional side on climate refugees—and there’s been far too little—has focused on helping people from island nations, because it’s so recognizable that sea level rise is an urgent threat there.

I know you’re still in the early stages of your work under the remit of the IGCC grant, but are there any interesting things that you’ve learned so far?

Kate Ricke: This is kind of boring, but what we’ve done so far under the remit of this grant is deal with a new dataset and new analytical approaches. We had previously been using a global migration dataset because it was the only thing available, but people weren’t too sure about whether it was a reliable dataset. In the last year, a new global-scale migration dataset with multiple timesteps was released, and so we’ve been spending a lot of time processing and validating that dataset and trying to figure out how to use it instead of the imperfect one we were using before. We’re really close to being able to apply it and feel good about using it.

We’ve had a lot of trouble figuring out a scientifically defensible way to choose the statistical model that we apply to understand what the migration response to a particular stressor would be and then project that into the future. To do this, we’ve been developing a machine learning-based framework for model selection. It’s called a “lasso method,” and it helps us have a way to defend why we selected the statistical model that we did. There are just so many ways that you could characterize climate change in terms of processing physical data, and it’s hard to pick which one is the right one or what’s closest to right, because of course, nothing’s exactly right. So that’s the other thing we’ve spent a lot of time on in the last year—building the algorithms to do scientifically defensible model selection.

Gaurav Khanna: Like Kate said, one of the components is making sense of what is important in determining migration. Is it changes in temperature or precipitation? Or is it the variability of temperature? Or is it the variability of temperature interacting with precipitation? Many of these variables might be correlated with each other. Maybe the variability in temperature is highly correlated with the change in temperature, so it might not make sense to include both of them in the same model, because one of them already captures the information that the other has. This machine learning process helps determine what the important variables are in a sparse way.

Social scientists and data scientists have been using the lasso method for all other kinds of problems, and we’re trying to apply it now to the climate science side. At this stage, we want to be a little agnostic about what the model is. We tell the machine what the candidate variables are—variability and changes in trends, and things like that—because we think those make sense. But then we let the data tell us what’s really important here.

What do you hope will be the implications of your project? What is your intended audience?

Gaurav Khanna: The dream is to reach a really wide audience. We do care a lot about the policy audience with this work. Of course, it has to be vetted by experts in the field, which means a review by other academics who’ve been deeply involved in this research. So, we first need their approval before we can engage with the policy audience on it, because we want to make sure that the experts in the field think what we’re doing is correct. Then we want to engage with the policy audience much more broadly.

Kate Ricke: For me, since this is kind of out of my main area of academic interest, if I didn’t think that we could eventually have some influence on policy, I don’t think I would have started this. But, in order to have influence on policy, we have to convince the other people who are experts on the topic. Because it’s such a huge policy challenge with such huge implications for human welfare, this is an area of climate change impact science and a place for adaptation planning that could make a huge difference if we can say something helpful.

Thumbnail credit: United Nations Photo